Expression of conjugation genes is controlled by a novel zipper-type transcriptional attenuation mechanism

Abstract

Proper expression of genes clustered in operons, particularly large operons, can be complex, often involving multiple regulatory switches. The conjugation operon present on the Bacillus subtilis conjugative plasmid pLS20 is over 32 kb long. This operon starts with a 456-nt leader region, which is followed by the first two genes of the operon, encoding a two-component processive antitermination system. Here, we demonstrate that the long leader region encodes a transcriptional attenuator that we named cATT_pLS20. In vivo and in vitro analyses showed that the attenuator is composed of three segments: a stem–loop structure with a long imperfect stem that is preceded by a sequence that may form a weak stem–loop and followed by a sequence that may form an intrinsic terminator. Sequences of the upstream stem–loop and downstream terminator are complementary, allowing further extension of the long stem, thereby generating an antiterminator conformation. Based on similarity with a zipper, we coined this the zipper-type attenuator. Similar zipper-type attenuators are present upstream of the first genes of conjugation operons of all pLS20 family plasmids, as well as on many other conjugative plasmids in Gram-positive bacteria, suggesting that a common attenuation mechanism regulates expression of many conjugative operons.

Graphical Abstract

Graphical Abstract.

Introduction

Gene expression is a highly dynamic process with only a portion of the genome expressed at any given time. The expression profile can change rapidly and profoundly in response to changing environmental conditions. Multiple mechanisms acting on the three stages of the transcription cycle—initiation, elongation, and termination—have evolved to achieve these adaptive gene expression profiles [1–11].

Bacterial conjugation is the process by which a genetic element, named the conjugative element, can transfer from a donor to a recipient cell via a channel that connects both cells [12–14]. Often a conjugative element is located on a plasmid, which is then called a conjugative plasmid. In addition to the metabolic burden, expression of conjugation genes significantly affects the structure and the physiology of the bacterial host. For instance, some of the conjugation genes encode cell wall-associated adhesins and integral membrane proteins that change the characteristics of the bacterial cell wall and membrane. The adhesins facilitate attachment of donor to recipient cells, while the membrane proteins organize into a membrane-embedded type IV secretion system (T4SS) that is responsible for transferring a ssDNA strand of the plasmid into the recipient cell [14–17]. The conjugation process also affects the biology of the plasmid such as the mode of DNA replication. At the onset of conjugation, the plasmid switches from the common theta replication mechanism used to duplicate the plasmid to a rolling-circle mechanism to generate the ssDNA strand that is transferred to the recipient cell. Since the onset of conjugation affects various aspects of the host and plasmid biology, expression of the conjugation genes are tightly controlled [18–20], which is the case for the Bacillus subtilis conjugative plasmid pLS20 [21]. This 64 kb plasmid contains a large operon of >32 kb that contains at least 38 genes numbered from gene 28 to 64. These genes, collectively called conjugation genes, encode the structural conjugation proteins as well as a protein and RNA that are involved in the processive antitermination (P-AT) mechanism (see below), and some proteins of unknown function. We previously identified regulatory mechanisms that control (i) the initiation of transcription at the single strong conjugation promoter P_c, located upstream of the first gene of the conjugation operon (gene 28a), and (ii) processive transcription of the extremely long conjugation operon.

Initiation of transcription at the conjugation promoter P_c involves multiple layers of control, involving three proteins encoded by genes 25–27 located immediately upstream of the conjugation operon. Gene 27, whose orientation is divergent to the genes in the conjugation operon, encodes a transcriptional repressor named Rco, while genes 25 and 26 form a bicistronic operon encoding Rap, the Rco antirepressor, and a small prepropeptide Phr, respectively. Rco tetramers can bind two operators, one of which overlaps the divergently oriented promoters P_c and P_r, with the latter driving expression of Rco (Fig. 1A). When bound to its operators, Rco activates expression of the weak P_r promoter and tightly represses the strong P_c promoter, keeping conjugation in the default ‘off’ state. Transcription from P_c is induced when Rap binds to and inactivates Rco. After secretion into the medium and a second proteolytic step, the pentapeptide Phr* corresponding to the five C-terminal residues of Phr is generated, which acts as a signal peptide. When imported by donor cells, Phr* interacts with Rap thereby inactivating its antirepressor function, and allowing Rco to reestablish repression of the conjugation operon. High Phr* levels are generated only when all or most cells in a population contain pLS20. Therefore, this quorum sensing mechanism allows expression of the conjugation genes only when donor cells are potentially surrounded by plasmid-free recipient cells. The three proteins Rap, Phr*, and Rco have been characterized in considerable detail at the functional, biochemical, biophysical, and structural levels [22–26].

Figure 1.

An intrinsic terminator (Ter_up28) precedes conAn1, the first gene of the conjugation operon (gene 28a). (A) Schematic diagram focusing on the intergenic region between the conjugation promoter P_c and the first gene of the conjugation operon, conAn1 (black arrow). The positions of the divergent promoters P_r and P_c are indicated with a yellow and blue box, respectively. Transcription start sites are indicated with vertical lines. Transcripts derived from P_r and P_c are indicated with yellow and red wavy lines, respectively. Numbering shown at the top corresponds to the P_c-derived transcript; position 1 is the transcription start site. Rco binding sites in the Rco operators OII and OI, are indicated with grey boxes labelled a and b, respectively [23]. The putative intrinsic terminator Ter_up28 is indicated. The sequence of the region encompassing Ter_up28 is shown. Sequences forming the predicted stem, loop, and U-rich tract of the terminator are shown in blue, purple, and red, respectively. Sequences that would form the stem are also marked with blue arrows. The likely Shine–Dalgarno (SD) sequence and the first two codons of conAn1 are indicated in green. In the upper panel, genes rco, rap, and phr, located upstream of the conjugation operon, are also indicated. Note that this region is not drawn to scale. (B) Predicted structure of Ter_up28. The predicted thermodynamic stability is indicated. The colour code to indicate sequences forming the stem, loop, and U-rich tract is the same as those used in panel (A). The position of the G to A point mutation in the 5′ stem of Ter_up28 present in strain DGA06 is indicated. The terminator was identified with the ARNold webserver [41, 42]. (C) DNA sequence and features of the P_c conjugation promoter and leader region. Promoter P_c with an extended −10 sequence, the transcription start site (+1), Rco binding sites (R-box), and the start of gene 28a (conAn1) with its SD sequence are indicated. Numbering corresponds to the P_c-derived transcript. The region between positions 230 and 410 contains four inverted repeat sequences: IR-1, IR-2, IR-3 and IR-4 (indicated with arrows). These IRs are predicted to form stem–loop (SL) structures in the transcript with calculated free energies at 37°C of −7.5 (SL-1), −6.4 (SL-2), −10.3 (SL-3), and −11.3 (SL-4) kcal/mol. SL-4 is the Ter_up28 hairpin. RNA structure predictions suggest that the SL derived from IR-3 forms the apex of an extended imperfect SL, which can be extended even further to include sequences of SL-2 and Ter_up28 (see text). These extensions are indicated with light blue and pink lines, respectively. Mismatches and bulges in these regions are indicated with thin lines. The position of the likely transcription termination site for Ter_up28, position 419, is shown in red. Sequences of the three oligonucleotides oligo 1, oligo 2, and oligo 3 used in the in vitro transcription experiments are indicated with black arrows containing dashed lines.

The first two genes of the conjugation operon, genes 28a and 28b, code for a two-component P-AT system, named conAn [27]. Gene 28a encodes protein ConAn1 and gene 28b encodes an RNA, named ConAn2. Whereas ConAn2 exerts antitermination, ConAn1 acts as a processivity factor. The conAn antitermination system functions only in cis. The association of ConAn1 and ConAn2 with RNA polymerase (RNAP) that has initiated transcription at P_c alters the elongation complex such that it is capable of bypassing at least 6, and probably up to 20 intrinsic terminators, that are distributed along the conjugation operon. A specialized intrinsic terminator with an exceptionally long stem that is resistant to conAn-mediated antitermination demarcates the end of the conjugation operon [28].

Here, we show that in addition to these mechanisms, the conjugation genes of pLS20 are controlled by a transcriptional attenuator located in the 456 nt leader region upstream of conAn1 (gene 28a). Thus, conjugation gene transcription is controlled by attenuation and P-AT mechanisms that act at a single and at multiple intrinsic terminators, respectively. Importantly, we found that a transcriptional attenuator similar to that on pLS20 precedes the conjugation operons of other pLS20 family plasmids, and many other plasmids replicating in Gram+ bacteria, suggesting that conjugation is commonly controlled by these dual post-transcription initiation control mechanisms.

Material and methods

Bacterial strains, plasmids, media, and oligonucleotides

Escherichia coli and B. subtilis strains were cultured in Lysogeny Broth (LB) medium [29]. Cultures were grown either in liquid medium with shaking or on 1.5% LB agar plates at 37°C. When required, antibiotics were added as follows: ampicillin (100 µg/ml) for E. coli; and spectinomycin (100 µg/ml) and chloramphenicol (5 µg/ml) for B. subtilis. All B. subtilis strains used in this study were isogenic with strain 168. A list of bacterial strains is provided in Supplementary Table S1. Plasmids and oligonucleotides are listed in Supplementary Tables S2 and S3, respectively. All oligonucleotides were synthesized by Integrated DNA Technologies (IDT).

Transformation and conjugation

E. coli competent cells were prepared and transformed using standard methods [30]. Generation of competent B. subtilis cells and transformations were prepared as described previously [31]. Conjugation experiments were performed at 37°C using a standard protocol [22]. Briefly, overnight cultures of donor and recipient cells were grown in LB medium in the presence of appropriate antibiotics. The next day the cultures were diluted 200-fold in fresh LB medium without antibiotic. At late exponential growth phase (OD₆₀₀ = 1), equal volumes (200 μl) of both cultures were mixed and incubated for 20 min without shaking, after which proper dilutions of the cells were spread on selective plates to determine the number of donor and transconjugants after overnight growth. The conjugation efficiency was expressed as the number of transconjugants per donor.

Construction of plasmids and strains

DNA techniques were performed using standard molecular methods [30]. Plasmids were isolated from E. coli strains using the NZYMiniprep kit (NZYtech). PCR (polymerase chain reaction) fragments were purified using NZYGelpure (NZYtech). In some cases, complementary primers were hybridized, and the resulting product was used as an insert in ligation reactions. In these cases, a 200 μl mixture was prepared (50 mM NaCl and 40 mM Tris–HCl, pH 7.5) containing 2000 pmol of each primer. After boiling, the solution was slowly cooled to room temperature. The hybridized oligonucleotides were then used as an insert in ligation reactions with plasmid pAND101 [27] digested with SalI and NheI. All enzymes used were purchased from New England Biolabs, USA (NEB). Cloned fragments were confirmed by DNA sequencing. The Q5 site-directed mutagenesis kit (NEB) was used for introducing point mutations in the attenuator region of strains DGA73, DGA84, and DGA86.

Specific information of the construction of each plasmid, including the oligonucleotides and restriction enzymes used to amplify and clone the fragments, is presented in Supplementary Table S2. Oligonucleotides used are listed in Supplementary Table S3. Derivatives of vector pAND101 were used to test terminator or promoter activity in vivo. First, a PCR fragment containing a putative terminator was cloned at the multiple cloning site located between the P_spank promoter and gfp on plasmid pAND101. The resulting plasmid was linearized and the P_spank-[insert]-gfp construct was integrated into the amyE locus by transformation. The resulting strains were used to study terminator or promoter activity of the cloned region by growing cells in the presence or absence of Isopropyl β-D-1-thiogalactopyranoside (IPTG), respectively.

Markerless in-frame deletions of pLS20cat were constructed making use of the pMiniMAD2 vector [32]. This method is based on single cross-over integration of a temperature-sensitive replicon via homologous recombination at the restrictive temperature, followed by permitting replication of the integrated replicon by growth at the permissive temperature to facilitate deletion of the replicon, thereby generating the desired deletion [33]. Briefly, ∼600 bp of DNA upstream and downstream of the region of interest were amplified by PCR, which were then joined in a second step by overlapping PCR. The resulting fused PCR product was purified, digested with NdeI and BamHI, and then cloned into pMiniMAD2 digested with the same enzymes. For a full description of the strategy see Meijer et al. [31]. More details regarding the construction of plasmids are described in Supplementary Tables S2 and S3.

Flow cytometry

Bacteria cultures were grown overnight at 37°C with shaking in LB medium and diluted 1:100 into fresh LB the following day. Cultures were incubated at 37°C with shaking until reaching an OD₆₀₀ of 0.8–1.0. Cells were harvested by centrifugation in 2-ml microcentrifuge tubes and washed twice with phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄). The final pellet was resuspended in 2 ml PBS. Fluorescence was measured using a FACSCalibur flow cytometer (Becton Dickinson, USA) equipped with a 488-nm argon laser. For each sample, at least 500 000 events were collected using a 530/30-nm band-pass filter. Data acquisition was performed with CellQuest Pro (Becton Dickinson), and analyses were carried out using FlowJo v6.4.1 (TreeStar, USA). Based on forward (FSC) and side scatter (SSC), a gating was applied for each experiment to select at least 100 000 cells having similar size and shape that reflect the majority of the population. Each experiment was repeated on different days until at least triplicates were obtained. For each individual experiment, the fluorescence value was calculated as the geometric mean (Geomean) fluorescence intensities obtained from ≥100 000 cells. Finally, for every strain the mean value of the replicates was calculated together with their standard deviation, which are represented in the Fig. 2A.

Figure 2.

A transcription attenuator precedes the first gene of the pLS20 conjugation operon. (A) Schematic diagram of the promoter and leader region of gene 28a encoding ConAn1. Inverted repeated sequences 1 to 4, which are predicted to form SL structures in the transcript, are indicated with the following colours: yellow (SL-1), green (SL-2), blue (SL-3), and red [SL-4 (Ter_up28)]. For other symbols and colours, see Fig. 1. (B) The top line shows features of the relevant region of AND101 spanning the P_spank promoter and gfp (green arrow). The lower panel provides a schematic overview of the different regions of the 5′ leader that are present in the corresponding integrated fusions. Information about the cloned fragments, their name (F), the corresponding strain name, and the fluorescence levels obtained are shown on the right. These levels are expressed relative to the expression levels obtained for the negative control strain AND101 lacking an insert and grown in the presence of 1 mM IPTG. The vertical red lines in fragments F1* and F6B-E reflect the approximate position of mutations in these fragments. The fluorescence levels are given as quantitative values, expressed as arbitrary units, relative to the fluorescence level obtained with the negative control strain AND101, indicated as R-fluorescence, as described previously [51]. (C) Predicted antitermination conformation of the 5′ leader region spanning positions 231–419. Sequences forming SL-1, SL-2, SL-3, and Ter_up28 (SL-4) in the predicted termination conformation are highlighted using the same background colours as in panels (A) and (B). Nucleotides that would form the stems of SL-2 and Ter_up28 in the terminator conformation are indicated with arrows. The U-rich tract of the terminator is indicated with a red dashed line. The three Gs that are mutated to Cs in fragment F6B are highlighted in blue. The C and G residues that are mutated in fragments F6C-E are highlighted in red. (D) Predicted secondary structures of the wild type (WT) sequence (left) and the triple mutant in fragment F6B (right) of SL-2. G to C mutations are highlighted in blue. (E) Predicted secondary structures of SL-2 mutation C270T (fragment F6C, left), G276C (fragment F6D, middle), and the double mutant C270T:G276C (fragment F6E, right). The calculated free energies of these structures are given at the bottom. Note that the mutations in fragment 6C-E, but not the ones in fragment 6B, are predicted to strengthen the secondary structure.

Bioinformatics analysis of sequences upstream of conjugation operons

The 35-member family of pLS20 plasmids [34] together with newly identified plasmids pBcaBIK3 and pBsuNRS6190 [35] were screened for the presence of a putative attenuator upstream of their conjugation operon. For this analysis, the 280 bp regions upstream of the gene 28a (conAn1) homologues were extracted and identical sequences were purged resulting in 19 unique sequences. These sequences were used to (i) generate a multiple sequence alignment (MSA) with LocARNA using standard settings [36], and (ii) determine the phylogenetic relationship by constructing a maximum-likelihood tree with IQ-TREE software [37] version 2.2.6 using the edge-linked model [38] and applying the following settings: automatic selection of model, 1000 bootstrapping, and 1000 replicates to perform an SH-like approximate likelihood ratio test.

The following strategy was used to identify putative attenuators on conjugative non-pLS20 family plasmids of Gram-positive bacteria. Plasmid sequences deposited in the plasmid database [39, 40] were screened for the presence of a gene 28a (conAn1) homologue that was followed downstream by typical conjugation genes like ATPase-encoding genes virB4 and virD4, resulting in 1162 hits. The 500 bp region immediately upstream of the gene 28a of each plasmid was extracted. After purging duplicates and removing sequences belonging to plasmids of the pLS20 family, 423 unique sequences were obtained. According to the ARNold webserver [41, 42], 339 of these sequences were predicted to contain a putative terminator (see Supplementary Table S4). Strikingly, in many cases the predicted terminator was located closely upstream of conAn1 (gene 28a), as in the case of the pLS20 family plasmids. The presence of a terminator upstream of the first gene of a long conjugation operon suggests it might function as an attenuator.

In vitro termination assays

In vitro transcription was performed by modifying published procedures [43–45]. A double-stranded gBlock gene fragment was designed as the template for multi-round in vitro transcription experiments. This fragment consisted of a strong engineered promoter and the pLS20 conjugative operon leader region from positions 199–466 with respect to the transcription start site (TSS). Primers were designed to PCR amplify the gBlock fragment to use as the template for in vitro transcription. The RNAP open complex was assembled by mixing equal volumes of 2× template (100 nM) and 2× RNAP master mix (100 μg/ml bovine serum albumin, 150 μg/ml B. subtilis RNAP holoenzyme, and 2× transcription buffer) and incubating it for 5 min at 37°C. Transcription was initiated by adding an equal volume of 4× elongation master mix containing 1.2 mM ATP, 1.2 mM GTP, 1.2 mM CTP, 0.4 mM UTP, 0.6 μCi of [α-³² P] UTP, 320 mM KCl, and 1× transcription buffer [40 mM Tris–HCl, pH 8.0, 5 mM MgCl₂, 4 mM dithiothreitol (DTT), 0.3 mM ethylenediaminetetraacetic acid (EDTA), and 5% glycerol]. The reaction mixture was incubated for 25 min at 37°C and transcription was stopped by adding an equal volume of 2× gel loading buffer [20 mM EDTA, pH 8.0, 0.025% sodium dodecyl sulphate (SDS), 0.025% bromophenol blue, and 0.025% xylene cyanol in formamide). To test the effects of intrinsic termination factors, NusA (4 μM), NusG (4 μM), or both were added to the open complex and incubated for 5 min at 37°C prior to adding the elongation master mix. DNA oligonucleotides were premixed in 10-, 100-, and 1000-fold molar excess over template with 4× elongation mix and added together when used. The samples were denatured by heating for 8 min at 80°C and fractionated on standard 5% polyacrylamide sequencing gels containing a 19:1 ratio of acrylamide:bis-acrylamide. RNA bands were visualized with a phosphorimager and quantified using Image-Quant software (GE Healthcare Life Sciences). Each in vitro transcription experiment was performed twice. Termination efficiency (%T) was calculated by quantifying the relative intensity of the terminated transcript to the readthrough transcripts that failed to terminate using the following equation: %T = (a/a + b) x 100, with a equal to the intensity of the terminated band and b equal to the intensity of all readthrough bands that failed to terminate. Since UTP was used as the labelling nucleotide, we also included a correction factor to take into consideration that the terminated and runoff bands contain 80 and 94 U residues, respectively.

RNA isolation and RNA sequencing

For total RNA extraction, bacterial cultures were grown overnight in liquid media with shaking at 37°C and diluted 1:100 in fresh LB media the next day. The fresh cultures were grown at 37°C with shaking until the culture reached an OD₆₀₀ = 0.8–1.0, at which time 1.5 ml of the culture was harvested by centrifugation and stored immediately at −80°C. Pelleted cells were thawed on ice and then lysed using a bead mill. Total RNA was isolated using the Monarch Total RNA Miniprep Kit (NEB) and stored at −80°C. The concentrations of RNA samples were measured using a NanoDrop spectrophotometer (Thermo Fisher Scientific), and 500 ng of RNA was fractionated through a 1% agarose bleach gel to assess the quality. The RiboCop rRNA depletion kit (Lexogen Vienna, Austria) was used to remove rRNA from 500 ng total RNA. Subsequently, the CORALL Total RNA-SeqTotal RNA Library Kit (Lexogen Vienna, Austria) was used to prepare libraries for Illumina sequencing. Samples were sequenced on the Illumina NextSeq 1000 to generate 100 nt single end reads (100SE) with an average read-depth of 8–12M reads per sample. The quality of the resulting fastq reads were checked using FastQC v0.11.9 (Babraham Bioinformatics, Cambridge) and mapped on the reference genome [B. subtilis subsp. subtilis strain 168 (GenBank identifier AL009126.3)] using Bowtie2 v2.4.2 using default settings [46]. The resulting SAM files were converted to BAM using SAMtools 1.11, and featuresCounts 2.0.1 was used to obtain the gene counts for each gene annotated in the reference genome, taking into account the orientation of genes and reads [47, 48]. Duplicate samples were summarized by calculating the mean counts for each gene. Bioinformatics analysis of RNA-seq data was done as described previously [22].

RNA structure measurement by SHAPE-MaP

Bacterial growth and cell collection

Bacillus subtilis strain PKS11 was grown in duplicate in LB medium at 37°C with shaking at 120 rpm until the culture reached an OD₆₀₀ = 0.6. For each culture, 1 ml was transferred into two microcentrifuge tubes corresponding to SHAPE-modified (+) and untreated (−) samples. Cells were pelleted by centrifugation at 10 000 rpm for 3 min at room temperature.

In vivo SHAPE modification

Cell pellets were resuspended in 90 μl pre-warmed PBS (37°C). SHAPE modification was performed by adding 10 μl of 1 M 2-aminopyridine-3-carboxylic acid imidazolide (2A3) to the positive samples giving a final concentration of 100 mM. Negative control samples received an additional 10 μl PBS. Samples were incubated at 37°C for 5 min. The reaction was quenched by adding 100 μl of 1 M DTT.

RNA extraction and quality assessment

Following SHAPE modification, cells were pelleted and resuspended in 200 μl nuclease-free water. Cell lysis was performed by adding 20 μl of 10% SDS, 150 μl acid phenol, 150 μl chloroform, and 100 μm glass beads. Cells were disrupted using a bead beater for 3 min. After phase separation, the aqueous phase was collected and washed with 400 μl chloroform and centrifuged. Subsequently, 180 μl of the aqueous phase was transferred to a fresh tube containing 450 μl Monarch StabiLyse DNA/RNA Buffer (New England Biolabs). RNA purification was completed using the Monarch Total RNA Miniprep Kit (New England Biolabs) according to the manufacturer’s instructions. RNA concentration was determined using a NanoDrop. RNA integrity was assessed by electrophoresis on a 1.5% agarose gel containing 1:10 000 CyberSafe DNA stain.

Reverse transcription and PCR amplification

Reverse transcription was performed using 3 μg of total RNA in a reaction containing 1 μl dNTP mix (10 mM each) and 2 μl reverse primer (10 μM; 5′-CTATTCCCTCCCTAAATTTTCAATC-3′). The reaction volume was adjusted to 11.5 μl with nuclease-free water. Samples were denatured at 70°C for 5 min and immediately cooled on ice. An 8.5 μl reverse-transcription master mix was added, consisting of 4 μl 5× RT buffer (250 mM Tris–HCl, pH 8.0; 375 mM KCl), 2 μl of 0.1 M DTT, 1 μl SuperScript II reverse transcriptase (Thermo Fisher Scientific), 0.5 μl SUPERase•In RNase inhibitor (Thermo Fisher Scientific), and 1 μl of 120 mM MnCl₂. Reverse transcription was carried out at 42°C for 90 min, followed by sequential incubations at 50°C, 55°C, and 60°C for 10 min each. Complementary DNA (cDNA) was purified using 1.8× volume NEBNext® purification beads and eluted in 10 μl nuclease-free water.

PCR amplification was performed using 5.5 μl purified cDNA, 1 μl forward primer (10 μM; 5′-CATACACTATCAAAGTGTGCTATC-3′), 1 μl reverse primer (10 μM), and 7.5 μl Q5 Master Mix (New England Biolabs). Amplification was conducted with an initial denaturation at 98°C for 30 s, followed by 25 cycles of 98°C for 10 s and 61°C for 15 s. A final extension was performed at 65°C for 5 min. The resulting 294-bp amplicon was purified using 1.8× volume NEBNext® purification beads and eluted in 20 μl of 0.1× TE buffer.

Library preparation and sequencing

For each sample, 50 ng of purified PCR product was used for library preparation with the NEBNext FS DNA Library Prep Kit according to the manufacturer’s protocol. Libraries were sequenced on an Illumina NextSeq 1000 using a 100-cycle P1 flow cell, generating ~5 million reads per sample.

SHAPE data analysis

Sequencing data were analysed using RNA Framework. SHAPE reactivities were calculated following the standard SHAPE-MaP workflow implemented in the software suite as described previously [49].

Results

A 50-bp region located just upstream of the conAn1 antitermination gene (gene 28a) encodes an intrinsic terminator

A schematic view of the features within the intergenic region between gene 27 (rco) and the first gene of the large conjugation operon of plasmid pLS20, gene 28a (conAn1), is shown in Fig. 1A. The positions of the conjugation promoter P_c, its TSS, and the binding sites for the transcriptional repressor Rco, which is encoded by gene 27 in the opposite direction, were identified previously [23]. conAn1 is preceded by a 456-bp 5′ leader region. The DNA sequence of P_c and the 5′ leader region is shown in Fig. 1C. Throughout the paper, the positions of the features within the 5′ leader region are numbered relative to the P_c TSS. RNA structure predictions using TransTermHP (transterm.cbcg.umd.edu) [50], as well as Erpin and RNAmotif on the ARNold web server (rna.igmors.u-psud.fr/toolbox) [41, 42], were performed to predict whether the 5′ leader contains RNA structures. These analyses identified a putative intrinsic terminator (Ter_up28) just upstream of the SD sequence of gene 28a (conAn1, Fig. 1). If this region between positions 389 and 419 of the 5′ leader encodes a terminator, transcripts derived from P_c would terminate upstream of the conjugation genes, resulting in reduced expression of the conjugation operon.

We used an in vivo expression system to test whether this region contains a functional terminator. The system, which is based on the B. subtilis integration vector pAND101, was used previously to identify and characterize the two-component antitermination system conAn of pLS20 [27, 28]. Briefly, the amyE integration cassette present on this vector contains the IPTG-inducible P_spank promoter that drives expression of a gfp reporter gene. pAND101 was used to construct the B. subtilis control strain AND101 containing a single copy of P_spank-gfp integrated at the amyE locus. When grown in the absence of IPTG, no GFP was produced in AND101 cells. But when the strain was grown in the presence of IPTG, GFP was produced and cells became fluorescent [27]. The fluorescence level of individual cells was measured by flow cytometry. The DNA regions to be tested for termination activity were inserted into this vector between P_spank and gfp, generating P_spank-X-gfp, where X is the cloned fragment. Each resulting transcriptional fusion was then integrated into the amyE locus of the B. subtilis chromosome. When grown in the presence of IPTG, the fluorescence levels of cells would be lower than those of the control strain AND101 if the cloned fragment encodes a functional terminator. Several different regions of the 5′ leader were tested. For simplicity, we refer to the cloned fragments using numbers as indicated in Fig. 2.

To test whether the predicted terminator Ter_up28 is functional in vivo, a 49-bp region encompassing Ter_up28 [positions 381–429, fragment 1 (F1), Fig. 2B] was cloned upstream of gfp in pAND101 to generate pDGA05. In addition, a variant containing a G to A point mutation in the stem of Ter_up28 (Fig. 1B) was also constructed (pDGA06). This fragment was named F1*. The two pDGA vectors were used to construct the B. subtilis strains DGA05 (P_spank-F1-gfp) and DGA06 (P_spank-F1*-gfp), which were then analysed for GFP expression together with strain AND101 (negative control) by flow cytometry. In agreement with earlier results [27], the control strain AND101 displayed a high level of GFP fluorescence when grown in the presence of IPTG and this value was set to 100%. Importantly, a low fluorescence level (6.9%) was obtained for strain DGA05, showing that the presence of Ter_up28 upstream of the gfp reporter strongly inhibited expression of gfp, implying that the insert encodes a functional terminator. The predicted structure of this terminator suggested that the point mutation present in strain DGA06 would disrupt the 8-bp stem and consequently affect the function of Ter_up28. This prediction was supported by the observation that the fluorescence level obtained for DGA06 was six-fold higher than that of strain DGA05, which contained the wild type (WT) sequence (Fig. 2B). Together, these results provide strong evidence that the 49-bp region immediately upstream of the first gene of the conjugation operon (conAn1, gene 28a) encodes an intrinsic terminator.

Ter_up28 functions as a transcription attenuator

As Ter_up28 is located immediately upstream of the first gene of the conjugation operon, we suspected that this operon would be regulated by a transcription attenuation mechanism in which an alternative antiterminator structure would compete with Ter_up28 hairpin formation. Consistent with this hypothesis, our previous RNA-seq results obtained when conjugation levels were at their maximum showed that transcripts starting at P_c bypassed Ter_up28 [22, 27, 51]. As a first approach to test this possibility we constructed additional strains, each containing a different fragment of the leader region in front of the gfp reporter. As expected, strains DGA04 and DGA19 containing the 5′ leader fragments lacking Ter_up28 (fragments F3 and F4) displayed high fluorescence levels (Fig. 2). Importantly, high fluorescence levels were also obtained for strain DGA03 containing fragment F2, corresponding to the entire leader region and including Ter_up28 (positions 1–445, Figs 1 and 2). This result suggested that transcription of the 5′ leader region preceding Ter_up28 inhibited terminator function in vivo. As explained in more detail below, the 3′ half of the leader transcript contains predicted Stem loop (SL) structures. We therefore tested whether this RNA segment was important for the putative attenuation mechanism. Strain DGA02 containing gfp fused to fragment 5 also displayed a high fluorescence level. These results provided compelling evidence that sequences in the 3′ half of the leader region are responsible for generating an antiterminator structure that competes with formation of Ter_up28 in vivo.

Formation of Ter_up28 is prevented by an antiterminator structure in the 5′ leader

Transcription attenuation mechanisms typically involve the formation of mutually exclusive antiterminator and terminator structures, with the antiterminator sequences preceding and overlapping the terminator. Generally, the conformation that is energetically and/or kinetically more favourable will form by default, while binding of a factor that can compensate for the less favourable structure or interfere with formation of the more favourable structure is required to generate the alternative structure [9]. Thus, identifying the default conformation is crucial for understanding the function of the attenuator. We used in vitro transcription experiments to examine termination in the 5′ leader. These in vitro experiments allowed us to eliminate the contribution of any factor that might bind to the nascent transcript in vivo and thereby favour the switch from the default to the alternative structure. A DNA template containing the 3′ half of the leader (positions 199–466) fused to a consensus σ^A-dependent promoter was used with purified B. subtilis RNAP. Transcription termination at Ter_up28 occurred at or near position 419. This conclusion is based on the relative migration position of the terminated transcripts compared to migration positions of RNA with known lengths run in parallel on preliminary gels. We found that the termination efficiency (%T) of Ter_up28 was only 3%, indicating that an antiterminator structure was the likely default conformation of the nascent transcript (Fig. 3B). The general transcription elongation factors NusA and NusG are known to stimulate intrinsic termination of several hundred intrinsic terminators in B. subtilis [43–45]. Thus, we tested whether NusA or NusG could stimulate formation of the terminator. However, as shown in Fig. 3B neither factor stimulated termination of Ter_up28.

Figure 3.

In vitro termination assays confirm intrinsic termination and the presence of an antiterminator structure. (A) Schematic view of the DNA fragment that was used as a template for multi-round in vitro transcription assays. The DNA fragment contains a consensus σ^A-dependent promoter sequence (P) followed by positions 199–466 of the 5′ leader of the conjugation operon transcript. (B) In vitro transcription assay carried out in the absence (–) or presence of NusA (A), NusG (G), or both (AG). Positions of terminated (Term) and run-off (RO) transcripts are marked. The expected sizes of the terminated and run-off transcripts are 221 and 268 bp, respectively. (C) In vitro transcription assays performed in the absence (–) or presence of DNA oligonucleotides complementary to 5′ leader positions 216–234 (oligo 1), 274–301 (oligo 2), or 380–399 (oligo 3). Oligos were added in 10-, 100-, or 1000-fold molar excess over template DNA. In the final three lanes, oligo 2 was held constant at 100-fold molar excess. Termination efficiency (%T) and standard deviations (±sd) are shown at the bottom of panels (B) and (C).

An ∼20 nt region located 100 nt upstream of Ter_up28 is crucial for antitermination

Analysis of the 5′ leader region of gene 28a revealed that the upstream half does not contain inverted repeats that would be capable of forming stable RNA hairpins. However, four SL structures were predicted in the downstream half of the leader region, with the Ter_up28 hairpin corresponding to SL-4 (positions 389–409) (Figs 2 and 3). This RNA hairpin is relatively stable with a calculated free energy of −11.8 kcal/mol. The other three SLs have free energies of −7.5 (SL-1, positions 236–265), −6.4 (SL-2, positions 268–294), and −10.3 (SL-3, positions 317–352), respectively. These SLs are indicated in Fig. 2, and their predicted structures are shown in Supplementary Fig. S1. To test whether structures corresponding to SL-1, SL-2, or SL-3 can compete with formation of Ter_up28, we constructed three additional gfp fusion strains containing successive deletions of fragment F5; fragment 5 was shown above to interfere with Ter_up28 function (strain DGA02, (Fig. 2B). Strain DGA01 (fragment F6A) containing SL-2, SL-3, and SL-4, but not SL-1, displayed high fluorescence levels. In contrast, low fluorescence levels were observed for strain DGA38 (fragment F7) lacking SL-1 and SL-2, and for strain DGA39 lacking SL-1, SL-2, and part of SL-3. These results suggested that SL-2 contains sequences required for antiterminator formation (Fig. 2).

The PASIFIC webserver (https://www.weizmann.ac.il/molgen/Sorek/PASIFIC/) is designed to predict alternative RNA structures of putative riboregulators [52]. When 5′ leader sequences extending from position 231 to the probable transcription termination site (TTS) at position 419 was examined, a terminator conformation was predicted that included Ter_up28 (left panel, Fig. 4B). In addition to Ter_up28, SL-1, SL-2, and SL-3 were predicted to form within this structure. To facilitate interpretation, the same colour code used in Fig. 2B was used here. Figure 4B also shows that the stem of SL-3 can be extended. The extended version of SL-3 (ext-SL-3) encompasses positions 305–383 (indicated with a purple ellipse, see also Supplementary Fig. S1). PASIFIC also predicted an alternative antiterminator conformation in which the 3′ strand of SL-2 pairs with the 5′ stem and loop sequences of Ter_up28 forming a further extended long SL, which concomitantly prevents formation of SL-2 and the terminator hairpin (right panel, Fig. 4B). Remarkably, the calculated minimal free energies of the terminator and the antiterminator conformations were similar, suggesting that little energy is required to switch from one conformation to the other.

Figure 4.

Alternative terminator and antiterminator conformations are predicted to form in the 3′ half of the gene 28a leader region. (A) Schematic view of the P_c promoter and the conAn1 5′ leader region. The ∼ 3′ half of the leader region that contains four SLs was analysed for predicted secondary structures using the PASIFIC webserver. SL3 forms the apex of a two-stage longer imperfect SL. (B) Terminator (left) and antiterminator (right) conformations predicted for the 5′ leader region spanning positions 231 to 419 by the PASIFIC webserver using ‘simple antiterminator’ settings. Colour codes to visualize the SLs and the terminator are the same as those used in Figs 2 and 3.

The RNAfold webserver (http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi) is designed to predict general local RNA secondary structures [53, 54]. Essentially the same structures as those predicted by PASIFIC were predicted by the RNAfold webserver when RNA sequences of the 3′ half of the leader region ending either at the probable TTS or 3-nt downstream were analysed (see Supplementary Fig. S2). Also in this case, similar minimal free energies were predicted for both structures, although in this case the antiterminator conformation was predicted to be somewhat more stable.

Together, the in silico analyses suggest that the downstream half of the 5′ leader can adopt two conformations having similar minimal free energies. One of these corresponds to a terminator conformation in which a long central SL structure (ext-SL-3) is flanked at the 5′ side by stem–loop SL-2 and at the 3′ side by terminator Ter_up28. In the antiterminator conformation, ext-SL-3 is further extended through pairing of SL-2 and Ter_up28 sequences, preventing formation of these two SLs.

The pLS20 conjugative operon is regulated by a transcription attenuation mechanism

We performed in vitro transcription and in vivo expression experiments to test our attenuation model. The in vitro transcription assays were performed with the same DNA template as described above, in the absence or presence of three DNA oligonucleotides that were complementary to three different RNA segments of the 5′ leader region (see Fig. 1C). In each case, experiments were performed with 10-, 100-, and 1000-fold molar excess of oligo over template DNA. Oligo 1 was complementary to positions 216–234, which is just upstream of SL-1. This oligo served as a control because hybridization of the oligo to the nascent RNA would not interfere with the formation of the putative antiterminator structure or Ter_up28. As shown above (Fig. 3B), in the absence of any oligo, only ~3% of transcription was terminated, and as expected, the presence of oligo 1 did not affect the termination efficiency (Fig. 3C). Oligo 3 was complementary to positions 380–399 such that hybridization would prevent formation of Ter_up28. By itself, this oligo did not affect the already low termination efficiency of Ter_up28. The third oligonucleotide tested, oligo 2, was complementary to positions 274–301. Hybridization of oligo 2 was designed to prevent formation of the predicted antiterminator, thereby promoting formation of the Ter_up28 hairpin. Inclusion of this oligo resulted in >90% termination. We also performed experiments with a constant level of oligo 2 (100-fold excess) with increasing concentrations of oligo 3. In this experiment, hybridization of oligo 2 would promote termination by interfering with formation of the antiterminator unless oligo 3 also hybridized to the terminator hairpin. Oligo 2 promoted termination in the presence of 10- and 100-fold excess of oligo 3, but this effect was eliminated at 1000-fold excess of oligo 3 (Fig. 3C).

To further test the attenuation model in vivo, we introduced mutations into fragment F6A in the context of the gfp reporter in pAND101 (Fig. 2B) that were predicted to favour formation of Ter_up28. Strain DGA20 contains the gfp reporter fusion in which three G to C point mutations were introduced at 5′ leader positions 282, 286, and 290. In the antitermination structure, these Gs anneal with Cs at positions 402, 398, and 394, respectively. The mutations would prevent formation of these base pairs, thereby favouring formation of Ter_up28 (Fig. 2C). Note that these three point mutations had only a small effect on the predicted strength of SL-2 (Fig. 2D). Figure 2B shows that GFP expression of this fusion (strain DGA20, fragment 6B) was low, confirming that these mutations favoured formation of Ter_up28 as predicted.

Finally, to asses experimentally whether the leader region adopts the predicted antiterminator conformation, we performed SHAPE-MaP analysis. Mutational profiling (MaP) was conducted on a 189-nt transcript that encompasses stem–loops SL-1–SL-4 (the 230–419 nt region of the leader, see Fig. 1). The resulting nucleotide reactivities were mapped onto the transcript sequence. High SHAPE reactivity was observed for the SL-2 region, indicating that these nucleotides were predominantly single-stranded and hence did not form a stable SL structure (Supplementary Fig. S3). This observation is consistent with the predicted antiterminator conformation. In contrast, nucleotides within the SL-1 and SL-3 regions displayed low SHAPE reactivity, indicative of base pairing and supporting the formation of these SL structures in vivo. Together, these results provide experimental evidence that the leader region predominantly adopts the antiterminator structure. SL-4 (Ter_up28) also displayed low SHAPE reactivity, suggesting formation of the terminator conformation. This result was expected because all readthrough transcripts can refold such that a mixture of both the antiterminator and terminator conformations exist. As these two structures have similar thermodynamic stabilities (Supplementary Fig. S2B), it is likely that these structures exist at comparable levels in vivo even though only a small fraction of the RNA represent authentic terminated transcripts.

According to the attenuation model, the switch from the terminator to the antiterminator conformation is due to the annealing of nucleotides between Ter_up28 and SL-2. Consequently, while in the terminator conformation both SL-2 and Ter_up28 are formed, formation of these structures is prevented in the antiterminator conformation (Fig. 4). SL-2 is weak (calculated free energy of 6.4 kcal/mol) due to two mismatches in its stem. We wondered whether strengthening SL-2 would be sufficient to stimulate formation of Ter_up28. The portion of SL-2 that is predicted to anneal with part of Ter_up28 within the antiterminator structure is limited to the 3′ sequence of the stem (Fig. 2C). We therefore constructed derivatives of fragment 6A containing single (fragments 6C and 6D) or double point mutations (fragment 6E) in the 5′ portion of the SL-2 stem that would strengthen its structure by removing the mismatches (Fig. 2E). Importantly, these substitutions would not interfere with the base pairing that is required for antiterminator formation. These substitutions strengthened SL-2 by ~4 and 8 kcal/mol for the single and double mutations, respectively (Fig. 2E). Compared to the WT fragment 6A, expression of gfp was reduced by ~65% for the strains containing the single mutations, and it was almost eliminated in the double mutant strain. These results indicate that formation or stabilization of SL-2 favours Ter_up28.

The results from both the in vitro and in vivo experiments support the following conclusions. First, the 5′ leader adopts the antiterminator conformation by default. Second, formation of the antiterminator is driven by annealing between sequences within SL-2 and Ter_up28. Third, both SL-2 and Ter_up28 are concomitantly formed in the terminator conformation. Finally, SL-2 is a weak SL and strengthening its secondary structure moderately by a single point mutation causes a substantial fraction of the RNA molecules to adopt the terminator conformation. Taken together, our results demonstrate that transcription attenuation regulates the conjugation operon on pLS20. We name this attenuator cATT_pLS20 (conjugation attenuator of pLS20).

Ext-SL-3 is an essential part of the cATT_pLS20 attenuator

According to the attenuation model, SL-2 and Ter_up28 are separated by ∼90 nt that are contained within ext-SL-3 (Fig. 4). As an initial approach to analyse whether this conspicuous feature is important for functionality of the attenuator, we predicted the RNA structure of the 5′ leader region (positions 196–445) lacking the sequences that would form ext-SL-3 (positions 299–370). Using the PASIFIC webserver we predicted that this region would preferentially adopt the terminator conformation (Supplementary Fig. S4).

To test the importance of the ext-SL-3 region in vivo, we constructed an additional reporter strain, DGA111, in which gfp is preceded by fragment F9 (see Fig. 2B) that contains positions 231–419, but lacking sequences forming ext-SL-3 (positions 299–376). As shown in Fig. 2B, fluorescence levels of DGA111 cells were five-fold lower than those obtained for the parental WT strain DGA02, indicating that ext-SL-3 is important for formation of the antiterminator and proper functioning of cATT_pLS20. In addition to the gfp reporter strain DGA111, we tested the effect of deleting the ext-SL-3 region by constructing the pLS20cat derivative pLS20catΔextSL3, in which the leader region spanning positions 299–370 were deleted. The conjugation efficiency of pLS20catΔextSL3 was almost 1000-fold lower than that of the WT pLS20cat (3.2E-6 ± 6.3E-7 and 2.7E-3 ± 1.0E-3, respectively). To determine whether this strong decrease in conjugation efficiency was due to altered transcription, we used RNA-seq to compare expression of the conjugation operons of plasmids pLS20cat and pLS20catΔextSL3. This deletion resulted in a large decrease in expression of almost all genes in the conjugation operon (Fig. 5). The sole exception was gene 61 that is most likely preceded by a strong promoter (unpublished results). Taken together, these results provide compelling evidence that ext-SL-3 plays an important role in proper function of cATT_pLS20.

Figure 5.

Heat map representation of the expression profiles of the conjugation operons of pLS20cat and derivatives lacking sequences coding for ext-SL-3 or Ter_up28. (A) Schematic view of the genes forming part of the large conjugation operon (28a to 64) and the three upstream-located genes 25, 26, and 27c. (B) Heat maps. Left lane (wt) shows the square root of expression levels of genes of the wild-type plasmid pLS20cat covering a range from 0 (white, lowest level) to 45 (dark green, highest level). The middle (ΔextSL3) and right (ΔTer) lanes show the expression of these genes of plasmids pLS20catΔextSL3 and pLS20catΔTer, respectively. For each gene of pLS20catΔextSL3 and pLS20catΔTer the expression is plotted as the differential expression level relative to the corresponding gene of the wild-type plasmid pLS20cat. Differential expression levels are presented as square root transformation of the normalized coverage on a scale covering a range from −45 to +45 using shades of blue and red for decreased and increased expression, respectively. White reflects no change in expression. The P-values of the differential expression levels are given in Supplementary Table S5. Differential expression levels are shown for genes in the conjugation operon genes 28 to 64 (rectangular box) and the three genes upstream and downstream of this operon. Gene numbers according to our nomenclature [22] are given on the right. RNA was isolated from cells harvested in late exponential phase when conjugation efficiency is at its maximum [22].

We also constructed another derivative of pLS20cat, pLS20catΔTer, in which we deleted 51 bp spanning Ter_up28. RNA-seq analysis revealed elevated expression levels of the conjugation genes of pLS20catΔTer (Fig. 5), indicating that in the WT plasmid a portion of the transcripts derived from P_c adopt the terminator configuration. However, despite higher expression levels of the conjugation genes, the conjugation efficiency of pLS20catΔTer was ~30-fold lower than that of the WT plasmid (8.8E-5 ± 3.5E-6 and 2.7E-3 ± 1.0E-3, respectively). Although we do not know why overexpression of the conjugation operon compromises conjugation efficiency, one possibility is that overexpression disturbs the correct stoichiometry of proteins involved in the conjugation process. Alternatively, overexpression likely increases the amount of conjugation proteins and at least some of them may impair steps of the conjugation process. Whatever the reason, this result suggests that expression of the conjugation operon is tightly regulated to achieve maximum conjugation efficiency.

Conjugation operons present on other plasmids of the pLS20 family are preceded by a cATT_pLS20-like transcriptional attenuator

We previously showed that pLS20 is the prototype of a family of conjugative plasmids present in bacilli that share a common genetic organization [34]. At the time of publication, this family consisted of 35 plasmids; some had complete plasmid sequences, while some were reconstructed partially from individual contigs. Two phylogenetic trees were constructed for these 35 plasmids. One was based on the DNA sequences of the origin of replication regions and the other on protein sequences encoded by eight conjugation and two other conserved genes. The two trees were similar, and according to this analysis these plasmids could be divided into four clades, of which clades 1 (harbouring pLS20) and 3 could be subdivided into three and two subclades, respectively. Screening public databases we identified two additional pLS20 family plasmids, harboured by Bacillus cabrialesii strain BIK3 [35] (accession code NZ_JAHKKH010000018) and B. subtilis strain NRS6190 (accession code OX419552.1), that both belong to clade 1. Based on established nomenclature [34], we name these plasmids pBcaBIK3 and pBsuNRS6190, respectively.

In silico approaches were used to analyse whether the conjugation operons present on other plasmids of the pLS20 family are preceded by a putative transcriptional attenuator. For this endeavour, we extracted the 280 bp region upstream of conAn1 of each plasmid, and then removed identical sequences, which resulted in 19 unique sequences. An MSA showed that there is considerable deviation between these 19 sequences (Supplementary Fig. S5A). Despite being a short region, the maximum likelihood tree constructed from these sequences (Supplementary Fig. S5B) was similar to the two previous trees based on the sequences of the origin regions and on 10 conserved proteins [34]. Subsequent bioinformatics analysis resulted in the identification of a putative intrinsic terminator immediately upstream of the conAn1 SD sequence in all these plasmids. An MSA limited to the terminator region is shown in Fig. 6.

Figure 6.

MSA of the DNA regions encoding the intrinsic terminator upstream of gene 28a (conAn1) present on pLS20 family plasmids. RNA sequences that would form the stem, loop, and U-rich tract are shown in blue, brown, and red, respectively. The stem of the terminator contains mismatches in three plasmids. These nucleotides are shown in black. The conAn1 start codon ‘UUG’ or ‘GUG’ is shown in green, and regions containing the SD sequences are highlighted with a grey background. Names of the plasmids and the clade they belong to are given at the left and right of the sequences, respectively.

The MSA of the terminator regions also revealed several other similarities and differences. First, the spacing from the terminator to the SD of the conAn1 gene varies between 0 and 26 bp. Second, the stem sequences of clade 2 plasmids are considerably different from those of plasmids belonging to the other clades. Finally, an A tract is present upstream of the terminator hairpin in clade 2 plasmids, and the ARNold server predicts that these terminators function as bidirectional terminators.

We then analysed these 19 sequences using the PASIFIC server, which revealed the following. First, RNA regions of all plasmids tested are predicted to adopt a terminator structure that is similar to the one in pLS20 (Supplementary Fig. S6). Thus, all RNA sequences are predicted to form an ext-SL-3 that is preceded by two SLs (SL-1 and SL-2) and followed by the terminator SL. Second, for 16 out of the 19 sequences, the predicted antiterminator conformation was similar to that of pLS20; i.e. the ext-SL-3 is prolonged through the annealing of sequences of the terminator and SL-2.

The complementary sequences that are involved in antitermination are shown in Supplementary Fig. S6. The complementary sequences of the antitermination region are highly conserved for plasmids of clades 1, 3, and 4 (Supplementary Fig. S7). It is worth noting that PASIFIC predicted a distinct antiterminator conformation for plasmid pBatNRS213 of clade 1B, and plasmids pBli4092 and pBliYNP2 of clade 3B. In these three cases, the 5′ strand sequences of the terminator stem are predicted to anneal with 3′ strand sequences of ext-SL-3 rather than with 3′ strand sequences of SL-2 (see Supplementary Fig. S6). However, as shown in Supplementary Fig. S7, the 3′ strand sequences of SL-2 of these three plasmids are nearly identical to those present in other plasmids of clades 1 and 3. This observation suggests that, like the other clade 1, 3, and 4 plasmids, antitermination of these three plasmids are likely driven by annealing of sequences of the terminator and SL-2. Supplementary Fig. S7 also shows that the complementary sequences forming the antiterminator region are different for plasmids of clade 2, which is consistent with the observation that the terminator sequence of these plasmids differ from plasmids of the other three clades (Fig. 6). Taken together, these data suggest that a transcriptional attenuator precedes the conjugation operon of all pLS20 family plasmids.

To test the hypothesis that an attenuator precedes the conjugation operon of all pLS20 family plasmids, we used the gfp reporter assay to analyse the leader region of two other pLS20-family plasmids. The first one was clade IV plasmid pBspNMCC4, which is most distantly related to the pLS20-harbouring clade I ([34], Supplementary Fig. S5B). The second plasmid was clade 1B plasmid pBatNRS213, for which it was predicted that formation of the antiterminator would involve hybridization between sequences of the 5′ strand of the terminator and the 3′ strand of ext-SL3. For each plasmid, two gfp-reporter strains were constructed using the same strategy as that used for the analysis of the leader region of pLS20 (see Fig. 2B). In one of these strains, gfp was preceded by a short region corresponding only to the predicted terminator, and in the other strain gfp was preceded by a sequence spanning the entire leader region. The promoter regions of pBspNMCC4 and pBatNRS213 have not been determined and they could not be identified with certainty due to sequence divergence with pLS20. Thus, to ensure the presence of the entire leader, we cloned the DNA fragment containing the entire rco–gene 28a intergenic region of each plasmid. The results obtained for these four strains are presented in Supplementary Fig. S8, which show that the leader regions of these two plasmids encode an attenuator that functions similarly as cATT_pLS20. When gfp was preceded by the sequence corresponding to only the terminator regions, low fluorescence levels in the same range as that obtained with the negative control strain AND101 were obtained in the absence of IPTG for strains MBV07 and DGA117. Low fluorescence levels were also obtained for these two strains when they were grown in the presence of IPTG, indicating that these fragments encode a functional terminator. When grown in the presence of IPTG, the fluorescence levels obtained for strains MBV21 and MBV08, which contain the rco-28a intergenic region of plasmid pBspNMCC4 and pBatNRS213, respectively, were 20-fold higher than those obtained for the control strain AND101. High fluorescence levels were also obtained for these two strains when grown in the absence of IPTG. These results indicate that (i) the cloned fragments contain their respective P_c promoter that is stronger than the fully induced P_spank promoter, and (ii) the terminator rarely forms for transcripts including the entire leader region upstream of the terminator.

Discussion

We determined that transcription of the pLS20 conjugation operon is regulated by a transcription attenuation mechanism involving overlapping intrinsic terminator (Ter_up28) and antiterminator structures. This attenuator, which we name cATT_pLS20, is located in the 5′ leader region preceding the first gene of the conjugation operon, gene 28a encoding the antiterminator protein ConAn1. The downstream half of the RNA leader region can form four SLs. We show that SL-2, SL-3, and SL-4 (SL-4 is the Ter_up28 hairpin) participate in the attenuation mechanism (Figs 2 and 3). Under the conditions tested, SL-1 is not essential for cATT_pLS20 attenuator function. However, it is possible that SL-1 plays a role by acting as an insulator to prevent RNA sequences within the 5′ half of the leader region from interfering with the attenuation mechanism.

In silico analysis predicted that SL-3 forms the apex of the longer ext-SL-3. In the terminator conformation, SL-2 and Ter_up28 are located near the 5′ and 3′ of the base of ext-SL-3, respectively (Fig. 4). Sequences of the 3′ stem of SL-2 and the 5′ stem and loop of Ter_up28 are complementary allowing the already long stem of ext-SL-3 to be extended further, thereby adopting the antiterminator conformation while simultaneously preventing formation of SL-2 and Ter_up28 (Fig. 4). Thus, formation of the antiterminator occurs via a zipper-type mechanism that begins with ext-SL-3, which is propagated until the antiterminator structure is fully formed. Consequently, ext-SL-3 is an important component of the attenuator. This view is supported by the following results. First, our in vitro transcription studies showed that the antiterminator structure forms by default and that disruption of the antiterminator with an oligonucleotide promotes termination (Fig. 3). Second, instead of the antiterminator, the terminator structure is formed in the absence of ext-SL-3 (Figs 2 and 5, and Supplementary Fig. S4). Numerous transcription attenuation mechanisms have been described, but we believe this is the first example of a zipper-type attenuator.

SL-2 has a pivotal role in the switch from the antiterminator to the terminator structure. SL-2 is the weakest of the four SLs (−6.4 kcal/mol), suggesting that its structure is dynamic and easily destabilized, which is supported by the SHAPE-MaP results. Mutations to SL-2 that eliminated one or both of the two mismatches in the stem strengthened the structure by −4 to −8 kcal/mol, and this caused a large decrease in expression, consistent with the mutations causing most of the transcripts to adopt the terminator conformation (Fig. 2). These results show that switching between the antitermination and the termination conformations requires little energy, which is consistent with the similar calculated free energies for the terminator and antiterminator conformations (Fig. 4 and Supplementary Fig. S2). SL-2 is located upstream of Ter_up28 and hence the SL-2 sequences are synthesized before those of Ter_28up. Thus, formation and/or stabilization of SL-2 promotes formation of the terminator because it prevents completion of the antiterminator. Interestingly, the MSA presented in Supplementary Fig. S5 shows that the SL-2 region, and particularly the 5′ stem sequence, is the most conserved region of the 19 leader sequences of pLS20 family plasmids. This observation suggests that in addition to its structure, the RNA sequence may serve another important function by recruiting an RNA binding protein that might stabilize SL-2, and thereby promote terminator formation. Alternatively, it is possible that a counter (antisense) transcript that interferes with antiterminator formation is responsible for promoting termination. This type of counter transcript mechanism is responsible for inducing termination in the attenuator region that controls expression of the conjugation operon in the Enterococcus faecalis plasmid pCF10 [55].

We previously showed that pLS20 is the prototype of a family of related plasmids that all contain a long conjugation operon that is regulated by a conAn-type P-AT system [34]. Bioinformatics analysis indicated that all pLS20 family plasmids contain a zipper-type transcriptional attenuator upstream of their conAn1 gene, which we verified experimentally for plasmids pBspNMCC4 (clade IV) and pBatNRS213 (clade 1B) (Supplementary Fig. S8). In addition to the pLS20 family plasmids, a conAn-type P-AT system is located at the start of the conjugation operon on >350 plasmids of Gram-positive bacteria [27]. A similar strategy that was used for the pLS20 family plasmids was employed to determine whether these plasmids also contain a putative attenuator preceding their conjugation operon. By screening the plasmid database [39, 40], we identified 1162 plasmids containing a conAn1 homologue followed by typical conjugation genes such as ATPase-encoding genes virB4 and virD4. After removing identical sequences and those corresponding to pLS20 family plasmids, 339 of the remaining 423 unique sequences were predicted to contain a terminator, which in most cases was located just upstream of the conAn1 gene. Additional analyses suggested that these terminators form part of a transcriptional attenuator, and in many cases these attenuators appear to be of the zipper-type, which is a particularly common arrangement for attenuators present on plasmids from lacticaseibacilli.

Our results from this and prior studies provide the following model of how numerous conjugation operons present on plasmids from Gram-positive bacteria are regulated (Fig. 7).

Figure 7.

Schematic overview of the different levels regulating transcription of pLS20 conjugation genes. (A) Genetic map of pLS20. The large conjugation operon spanning genes 28a to 64 is indicated with a curved bar. The first and last gene of the conjugation operon are indicated (genes 28a and 64). Genes 25–27, located upstream of the conjugation operon encode the proteins regulating activity of the main conjugation promoter P_c. The attenuator cATT is shown as a blue lollipop symbol upstream of gene 28a. The multiple transcriptional terminators present in the conjugation operon are indicated with small red lollipop symbols. The large lollipop symbol labelled Ter_R located downstream gene 64 corresponds to the terminator that is resistant to the conAn P-AT system. (B) Proteins Rco (red), Rap (green), and Phr* (green), encoded by genes 27, 25, and 26, respectively, regulate transcription initiation. The main conjugation promoter P_c is repressed by default by the transcriptional regulator Rco. The antirepressor Rap is necessary to relieve repression. The activity of Rap is controlled by the quorum sensing peptide Phr*. See text for details. (C) Two mechanisms that regulate expression of the conjugation genes at the post initiation level: transcriptional attenuation (a, upper panel) and P-AT (b, lower panel). Transcriptional attenuation. A transcriptional zipper-type attenuator is able to adopt one of two mutually exclusive conformations preceding the first gene of the conjugation operon, gene 28a. The terminator is not formed in the antiterminator conformation, allowing RNAP (not shown) to proceed. The terminator is formed in the terminator conformation causing transcription to terminate before reaching the first conjugation gene. Processive antitermination. The first two genes of the conjugation operon, genes 28a (green arrow labelled 28a and conAn1) and 28b (purple rectangle, labelled conAn2), encode the protein and RNA component, respectively, of the two-component conAn-type P-AT system. Association of ConAn1/2 with RNAP changes its features allowing it to read-through any transcriptional terminators, except a terminator with special features, Ter_R, located at the end of the conjugation operon. See text for details.

Expression of the conjugation operons is repressed by default, and a complex multi-tiered regulatory system enables activation of the conjugation process under suitable conditions. This multi-tiered system includes at least three distinct levels. The first level regulates transcription initiation via a quorum sensing signal peptide system that controls the activity of the main conjugation promoter through derepression (Fig. 7B, see also the ‘Introduction’ section). The other two levels act during transcription elongation (Fig. 7C). The first of these two layers comprises the zipper-type attenuator described here (upper panel, Fig. 7C). At least in pLS20, pBatNRS213, and pBspNMCC4, the attenuator adopts the antiterminator conformation by default permitting transcription to proceed. The attenuator may function as a checkpoint mechanism to terminate transcription having started at the main conjugation promoter P_c. There are two non-mutually exclusive circumstances when this may occur; both of which are related to the way the main conjugation promoter is regulated. The multiple layers that control the initiation step of transcription at the P_c promoter ultimately depend on the level of the signal peptide Phr* (see the ‘Introduction’ section). Briefly, based on our current knowledge, the conjugation genes are transcribed when functional anti-repressor, Rap, is present to prevent the transcriptional repressor Rco from binding to the P_c promoter. Rap becomes inactivated upon interaction with the signal peptide Phr*, which effectively occurs when quorum levels of Phr* are reached. This regulatory switch implies that the conjugation process is activated whenever Phr* concentration is below the threshold level, regardless of other environmental conditions or physiological states of the donor cell where conjugation might not be beneficial. Thus, the attenuator may have evolved to inhibit conjugation by preventing the expression of the conjugation genes when Phr* levels are low, but other conditions are unfavourable to support the conjugation process. In addition or alternatively, the attenuator may function to ensure that proper amounts of the conjugation proteins are synthesized. Appropriate amounts of the conjugation proteins are probably needed for the conjugation process to occur properly. For instance, proper regulation of the synthesis of ssDNA that is transferred into the recipient cell may be disrupted when too many proteins involved in this process, named relaxosome proteins, are produced, and/or this may interfere with recruitment of the relaxosome complex to the T4SS. In addition, too many proteins of the T4SS may hamper proper formation of the T4SS or generate multiple T4SSs. In fact, deletion of the attenuator resulted in overexpression of the conjugation genes (Fig. 5) and this affected efficient conjugation. These observations are therefore in line with the view that the attenuator functions to modulate the expression level of the conjugation genes. In this respect it is important to consider that Rco not only represses the strong P_c promoter but also activates its own weak promoter P_r that is divergently oriented to P_c [22, 23, 51]. Hence, Rap-mediated removal of Rco from the DNA region containing these two promoters activates the P_c promoter and simultaneously inhibits de novo Rco synthesis. Thus, reestablishment of P_c repression depends primarily on the release of Rco from Rap–Rco complexes [22, 24, 25]. The Rco-Rap-Phr*-based regulatory switch may not always repress the strong P_c promoter in time to avoid generating too many transcripts and conjugation proteins and return to the ‘conjugation off’ state. Hence, the attenuator may have evolved as a negative feedback loop to modulate the expression level of the conjugation genes. This might for instance be achieved if the conjugation operon encodes a component stimulating the attenuator to adopt the terminator configuration, for example by stabilizing SL-2.

The second layer acting during transcription elongation is the conAn-type P-AT system (lower panel, Fig. 7C). Contrary to the zipper-type attenuator that acts locally by controlling a single terminator, the P-AT system allows the transcription elongation complex to bypass multiple intrinsic terminators. In addition to regulating transcription processivity, the P-AT system minimizes the effects of spurious transcription and allows differential expression of subsets of genes within the conjugation operon [27]. Thus, tight control of conjugation is achieved to prevent this energetically costly process from occurring in the absence of appropriate recipient cells and under otherwise unfavourable conditions.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by the Ministry of Economy and Competitiveness of the Spanish Government [Grant number PID2019−108778GB-C21 (AEI/FEDER, EU) to W.J.J.M.]; the National Institutes of Health (Grant numbers GM098399, GM153190 to P.B.); the Spanish Ministry of Universities and the Autonomous University of Madrid (Grant number CA1/RSUE/2021-00513 to A.M.-A.); L.J.W’s work is funded by a UK BBSRC Research Grant (BB/Y002644/1) awarded to R. Daniel and L.J.W.; and institutional funding from the Fundación Ramón Areces and Banco de Santander to the Centro de Biología Molecular Severo Ochoa. Funding to pay the Open Access publication charges for this article was provided by the CSIC Open Access Publication Support Initiative through its Unit of Information Resources for Research (URICI).

Data availability

RNA-seq and SHAPE MaP data were deposited to the NCBI SRA repository [Projects PRJNA1364995 and PRJEB115765 (ERP195964), respectively]. Flow Cytometry data are available from the corresponding author upon request. All other relevant data are given in the main text.